Search papers, labs, and topics across Lattice.
Guangdong Laboratory of AI and Digital Economy (SZ), China
1
0
3
4
MLLMs can ace the test, but still fail to *see*鈥攖hey often succeed at complex reasoning with symbols while failing at basic symbol recognition, revealing a reliance on linguistic priors over true visual perception.
